illustrious ULtraReal (Beta)
Full FP32 Model (Full Checkpoint)
Using FP32 CLIP is recommended but not required
Forcing FP32 UNET is not recommended for most users (BF16 UNET is recommend)
If using GGUF use these CLIP models
BF16 UNET and FP32 CLIP will see no slow down other then model load time for most users.
Note: I have multiple request to do illustrious, I always thought of it as a artist/art style model. I feel most of the function is gone when forcing the model back to realism.
Description
FAQ
Comments (7)
I tried to use the GGUF. Forge doesn't seem to know what the model is. I tried using with the Clip files and nothing seems to work. Maybe I am tired and missing something. Any help would be appreciated as this looks to be a decent model and would like to test it out.
COMFY is the workflow that I use for GGUF, however Friday the FP8 model is scheduled to release and should work with forge with no issues
@Felldude if you want to use Comfy without it getting too overwhelming, I recommend using SwarmUI. it's comfy, but auto-manages workflows and nodes based on your generation settings, and has a simple interface like forge.
Why there is no regular fp16 just like all the other models?!
If run with normal launch commands it would be downcast to FP16 by default
@Felldude Sorry but how do i use it correctly? Launching comfy with "--fp32-text-enc --fp32-vae --bf16-unet" as you mention on another model?
@Kerstal That is my general recommendation for COMFY UI, BF16 was recommended for FLUX and SD 3.5 and seems to work well with most of my FP32 trainings